Dr Roberto Bondesan is a Senior Lecturer in the Department of Computing. In this blog post, he shares more about his research as part of QuEST (Centre for Quantum Engineering, Science and Technology at Imperial College London). Roberto’s research explores the application of machine learning techniques to optimise the resources used by a quantum computer.
Can you tell us about your research area?
My research area is quantum computing. Quantum computers manipulate quantum information encoded in qubits to solve a computational problem and promise to enable the discovery of novel materials and molecules and to revolutionise cryptography. Qubits are implemented at the atomic level, where unwanted interference from the surrounding atoms is inevitable, and building a large-scale quantum computer capable of delivering computational advantages is one of the greatest technological and scientific challenges of our time.
My research focuses on the application of machine learning techniques to optimise the resources used by a quantum computer.
One of the main use cases for quantum computers is the simulation of quantum many body systems, and I am also interested in developing classical algorithms for benchmarking quantum algorithms for this task.
What led you to focus on this area?
My path to academia and quantum computing is a non-standard one. I did my PhD and postdoc in theoretical condensed matter physics, where I studied disordered and topological phases of quantum matter and their application to quantum computing. At the end of my postdoc, I got interested in machine learning and left academia for a research job in a tech company. In industry, I spent a few years applying machine learning to the optimisation of classical computer chips. My move back to academia was motivated by the will to put together my passion for quantum physics with that for machine learning to tackle important challenges that can benefit society.
What are the main aims of your current research?
One of the current aims of my research is to devise machine learning models to represent classically a quantum state. I am interested in the case of quantum many body systems where the exponential growth of the number of states with the number of qubits makes the problem particularly challenging. This problem is motivated by the application of quantum computing to simulate materials and molecules, where learning a classical representation from the outputs of the quantum computer can give important savings in the runtime of the quantum machine and enable downstream applications to material and molecule discovery.
Machine learning models that represent ground or thermal states can also be learned without quantum data—roughly speaking, by minimising the variational energy for a given Hamiltonian—and I am also currently working on using these methods as simulation methods for quantum many body systems.
Another aim of my research is to improve the efficiency of quantum error-correcting codes. A quantum error correcting code is a necessary component of a quantum computer that removes the noise in the computation. Quantum error correction requires a classical decoding algorithm that processes the data produced by the quantum computer to infer the error that occurred. This process is computationally challenging, and my aim is to develop a neural network-based decoder that enables real-time quantum error correction.
How could this research potentially benefit society?
We believe that a quantum computer would allow us to simulate molecules and materials accurately, which in turn can be used, for example, to develop better catalysts to store renewable energy and better drugs to cure disease. Quantum computers will also offer novel algorithms to solve hard optimisation problems that can be useful in many domains, such as making better policy decisions. These advances can benefit society at large. I am also a passionate advocate of using quantum computing for good.
What are the next steps in your research? Are there any challenges ahead?
One of the things I am most excited about in the future is to research novel quantum codes that can reduce the overhead of quantum error correction. This overhead comes from the fact that in error correction we encode information redundantly to protect it from noise. It has been recently realised that there exist families of quantum codes that can reduce this overhead dramatically, thus shortening the time to build a useful quantum computer. The price to pay for these new error correction schemes is, however, a much more challenging decoding problem. Currently, there are no good decoding algorithms known for this task, and this is one of the future problems I want to tackle.
Another topic I want to work on is the development of quantum algorithms for simulating quantum systems at finite temperatures. These algorithms are the quantum analogues of classical Monte Carlo algorithms, which are the main tool for studying classical systems in thermal equilibrium. The lack of a quantum computer to test the quantum algorithms on will require a lot of ingenuity to identify what problems they can solve better than a classical computer.